forest fire prevetion

7/31/2019 FOrest Fire Prevetion

1/17

Forest Fire prevention Using Wireless Sensor Networks 2011-12

CHAPTER- I

INTRODUCTION

The happenings of Forest Fire troubled every country all along, which caused great

losses every year. There was a message from China News Agency Beijing which was

reported on February 16, 2009. It said that the constant high temperature and drought, as

well as the precipitation decreased ,resulted in the high forest fire weather rating in some

provinces and regions of southern China. In addition, the growth of the moor bums which

was brought about by farmers lead to the number of forest fire increased rapidly. Forest fire

causes the timber to reduce the quantity and quality, so that a large number of animal and

plant deaths, destruction of nature and ecological balance, even small climate change of theforest. Above all, the prevention of the occurrence of forest fires is significant. Forest fires,

also known as wild fires, are uncontrolled fires occurring in wild areas and cause significant

damage to natural and human resources. Forest fires eradicate forests, burn the

infrastructure, and may result in high human death toll near urban areas. Common causes of

forest fires include lightning, human carelessness, and exposure of fuel to extreme heat and

aridity. It is known that in some cases fires are part of the forest ecosystem and they are

important to the life cycle of indigenous habitats. However, in most cases, the damage caused

by fires to public safety and natural resources is intolerable and early detection and

suppression of fires deem crucial. For example consider the resent incidents of forest fire in

Karnataka which is shown in Table.1.

Table. 1 Recent Forest fires in Karnataka

Forest Name Quantity of forest loss in hectares

Nagarahole(Hunasuru) 1400

Shettihalli(Shivamoga) 470

Arabbitthittu(Mysore) Not measured

Tharachegudde(Chikkamagaluru) 6

B.C Cavell(Turuvekere) 25

Kushalanagara(Kodagu) 500

Biligiriranganabetta(Chamarajanagar) 20

Shiradi 80

As the forest minister mentioned, the amount of forest lost in karantaka from past one

month (report is made on 29/03/2012) is 4000 hectares.

This technology presents a new real-time forest fire detection method by using

wireless sensor networks. The goal is to detect and predict forest fire promptly and

Dept of ECE., CIT Gubbi. 1


2/17


accurately in order to minimize the loss of forests, wild animals, and people in the forest

fire. In this proposed paradigm, a large number of sensor nodes are densely deployed in a

forest. Sensor nodes collect measured data (e.g., temperature, relative humidity) and send to

their respective cluster nodes that collaboratively process the data by constructing a neural

network. The neural network takes the measured data as input to produce weather index,which measures the likelihood for the weather to cause a fire. Cluster headers will send

weather indexes to a manager node via the sink. Then the manager node concludes the

forest fire danger rate based on received weather indexes and some other factors. In certain

emergent situations, sensor nodes may detect smoke or abnormal temperature. They will

directly send an emergence report to the manager node. To the best knowledge, there exists

no previous work that has proposed in-network processing using neural network in wireless

sensor networks.

This report exploits the essence of neural network that complex data processing can

be done by performing simple calculations at many organized single nodes. Such an

approach fits sensor networks well because individual sensor nodes have limited computing

capability. Moreover, the constructed neural network operates on vast raw data and extracts

small amount of information useful for the final decision; thus, both communication

overhead and energy consumption are significantly reduced. In addition to real-time forest

fire detection, the proposed sensor network approach can forecast potential forest fires, and

provide helpful information to extinguish forest fires and investigate the cause of the fires.

These features make the proposed paradigm superior to the traditional satellite-based forest

fire detection approach. Apart from preventive measures, early detection and suppression of

fires is the only way to minimize the damage and casualties. Systems for early detection of

forest fires have evolved over the past decades based on advances in related technologies. It

summarizes the evolution in the following, motivating the need and potential of wireless

sensor networks for this critical application.

1.1 Evolution of Forest Fire Detection Systems

Traditionally, forest fires have been detected using fire lookout towers located at

high points. A fire lookout tower houses a person whose duty is to look for fires using

special devices such as Osborne fire finder [Fleming and Robertson 2003]. Osborne fire

finder is comprised of a topographic map printed on a disk with graduated rim. A pointer

aimed at the fire determines the location and the direction of the fire. Once the fire location



3/17


is determined, the fire lookout alerts fire fighting crew. Fire lookout towers are still in use

in many countries around the world including USA, Australia, and Canada [B.C. Fire

Lookout Towers].

Unreliability of human observations in addition to the difficult life conditions for

fire lookout personnel have led to the development of automatic video surveillance systems

[Fire Watch Web Page; Breejen et al. 1998; Khrt et al. 2001]. Most systems use Charge-

Coupled Device (CCD) cameras and Infrared (IR) detectors installed on top of towers.

CCD cameras use image sensors which contain an array of light sensitive capacitors

or photodiodes. In case of fire or smoke activity, the system alerts local fire departments,

residents, and industries. Current automatic video surveillance systems used in Germany,

Canada, and Russia are capable of scanning a circular range of10

km in less than8

minutes. The accuracy of these systems is largely affected by weather conditions such as

clouds, light reflection, and smoke from industrial activities. Automatic video surveillance

systems cannot be applied to large forest fields easily and cost effectively, thus for large

forest areas either aeroplanes or Unmanned Aerial Vehicles (UAV) are used to monitor

forests. Aeroplanes fly over forests and the pilot alerts the base station in case of fire or

smoke activity. UAVs, on the other hand, carry both video and infrared cameras and

transmit the collected data to a base station on the ground that could be up to 50 km away.

UAVs can stay atop for several hours and are commanded by programming or joystick

controls.

More advanced forest fire detection systems are based on satellite imagery.

Advanced Very High Resolution Radiometer was launched by National Oceanic and

Atmospheric Administration (NOAA) in 1998 to monitor clouds and thermal emission of

the Earth. Moderate Resolution Imaging Spectroradiometer was launched by NASA. 1999

on board of the Aqua satellite to capture cloud dynamics.

Current satellite-based forest fire detection systems use data from these instruments

for forest fire surveillance. The instruments provide a complete image of the

earth every 1 to 2 days. The minimum detectable fire size is 0.1 hectare, and the fire

location accuracy is 1 km [L et al. 2000; Lohi et al. 1999]. The accuracy and reliability of

satellite-based systems are largely impacted by weather conditions. Clouds and rain absorb



4/17


parts of the frequency spectrum and reduce spectral resolution of satellite imagery which

consequently affects the detection accuracy.

Although satellite-based systems can monitor a large area, relatively low resolution

of satellite imagery means a fire can be detected only after it has grown large. More

importantly, the long scan period-which can be as long as 2 days-indicates that such

systems cannot provide timely detection. To summarize, the most critical issue in a forest

fire detection system is immediate response in order to minimize the scale of the disaster.

This requires constant surveillance of the forest area. Current medium and large-scale fire

surveillance systems do not accomplish timely detection due to low resolution and long

period of scan. Therefore, there is a need for a scalable solution that can provide real time

fire detection with high accuracy. It is believed that wireless sensor networks (WSN) can

potentially provide such solution. Recent advances in WSN support the belief that they

make a promising framework for building near real-time forest fire detection systems.

Currently sensing modules can sense a variety of phenomena including temperature,

relative humidity, and smoke [Crossbow Inc. Web Page] which are all helpful for fire

detection systems. Sensor nodes can operate for months on a pair of AA batteries to provide

constant monitoring during the fire season. Moreover, recent protocols make sensor nodes

capable of organizing themselves into a self configuring network, thus removing the

overhead of manual setup. Large-scale wireless sensor networks can be easily deployed

using aeroplanes at a low cost compared to the damages and loss of properties caused by

forest fires.

1.2 Contributions and Paper Organization

The presented report designs and evaluates the wireless sensor network for early

detection of forest fires. The design is based on solid forestry research conducted by the

Canadian Forest Service [Canadian Forest Service (CFS) Web Page ] over several decades.

In particular the contributions can be summarized as follows:

1. This report presents the key aspects in modeling forest fires, describes the Fire

Weather Index System [Canadian Forest Service (CFS) Web Page; de Groot 1998],

and shows how its different components can be used in designing efficient

fire detection systems. This could be of interest to researchers working in this areaand to sensor manufacturers who can optimize the communication and sensing

modules of sensors to fit forest fire detection systems.



5/17


2. Presents a simple data aggregation scheme based on the FWI System, which

significantly prolongs the network lifetime.

3. Shows how it can be extended to address several issues related to the forest fire

detection systems, such as providing different coverage degrees at the different sub

areas of the forest. This is important because, for example, the part of the forest thatare near to the residential areas need to be monitored with higher accuracy than

others.

CHAPTER-2

UNDERSTANDING AND MODELING FOREST FIRES



6/17


Forests cover large areas of the earth and are often home to many animal and plant

species. They function as soil conserver and play an important role in the carbon dioxide

cycle. To assess the possibility of fires starting in forests and rate by which they spread, it

uses the Fire Weather Index (FWI) System developed by the Canadian Forest Service

(CFS)[Canadian Forest Service (CFS) Web Page ], which is based on several decades offorestry research [San-Miguel-Ayanz et al. 2003].

The FWI System estimates the moisture content of three different fuel classes

using weather observations. These estimates are then used to generate a set of indicators

showing fire ignition potential, fire intensity, and fuel consumption. The daily observations

include temperature, relative humidity, wind speed, and 24-hour accumulated

precipitation, all recorded at noon Local Standard Time (LST). The system predicts the

peak fire danger potential at 4:00 pm LST. Air temperature influences the drying of fuels

and thus affects the heating of fuels to ignition temperature. Relative humidity shows the

amount of moisture in the air. Effectively, a higher value means slower drying of fuels

since fuels will absorb moisture from the air. Wind speed is an important factor in

determining fire spread for two main reasons: (a) it controls combustion by affecting the

rate of oxygen supply to the burning fuel, and (b) it tilts the flames forward, causing the

unburned fuel to be heated [Pearce 2000]. The last factor, precipitation, plays an important

role in wetting fuels.

Figure 2.1: Structure of Fire Weather Index (FWI) System.

As shown in Figure 2.1, the FWI System is comprised of six components: three fuel

codes and three fire indexes. The three fuel codes represent the moisture content of the



7/17


organic soil layers of forest floor, whereas the three fire indexes describe the behavior of

fire. In the following two sections briefly describe these codes and indexes.

2.1 Fuel Codes of the FWI System

The forest soil can be divided into five different layers [Canadian Forest Service

(CFS) Web Page; de Groot 1998] as shown in Figure 2.1 Each layer has specific

characteristics and provides different types of fuelsfor forest fires. These characteristics arereflected in fuel codes of the FWI System. Related to each fuel type, there is a drying rate at

which the fuel loses moisture. This drying rate, called time lag, is the time required for the

fuel to lose two-thirds of its moisture content with a noon temperature reading of 21C,

relative humidity of 45%, and a wind speed of 13 km/h [de Groot 1998]. Also, each fuel

type has a fuel loading metric, which describes the average amount (in tonnes) of that fuel

which exists per hectare.

Figure.2.2 Forest soil layers.

There are three fuel codes in the FWI System: Fine Fuel Moisture Code (FFMC),

Duff Moisture Code (DMC), and Drought Code (DC). FFMC represents the moisture

content of litter and fine fuels, 12 cm deep, with a typical fuel loading of about 5 tonnesper hectare. The timelag for FFMC fuels is 16 hours. Since fires usually start and spread in

fine fuels[de Groot 1998], FFMC can be used to indicate ease of ignition, or ignition

probability.

The Duff Moisture Code (DMC) represents the moisture content of loosely

compacted, decomposing organic matter, 510 cm deep, with a fuel loading of about 50

tonnes per hectare. DMC is affected by precipitation, temperature and relative humidity.

Because these fuels are below the forest floor surface, wind speed does not affect the fuel



8/17


moisture content. DMC fuels have a slower drying rate than FFMC fuels, with a time lag of

12 days. Although the DMC has an open-ended scale, the highest probable value is about

150[de Groot 1998]. The DMC determines the probability of fire ignition due to lightning

and also shows the rate of fuel consumption in moderate depth layers.

The last fuel moisture code, the Drought Code (DC), is an indicator of the moisture

content of the deep layer of compacted organic matter, 1020 cm deep, with a fuel loading

of about 440 tonnes per hectare. Temperature and precipitation affect the DC, but wind

speed and relative humidity do not have any effect on it due to the depth of this fuel layer.

DC fuels have a very slow drying rate, with a timelag of 52 days. The DC is indicative of

long-term moisture conditions, determines fires resistance to extinguishing, and indicates

fuel consumption in deep layers. The DC scale is also open-ended, although the maximum

probable value is about 800[de Groot 1998].

2.2 Fire Indexes of the FWI System

Fire indexes of the FWI System describe the spread and intensity of fires. There are

three fire indexes: Initial Spread Index (ISI), Buildup Index (BUI), and Fire Weather Index

(FWI). As indicated by Figure 2.1, ISI and BUI are intermediate indexes and are used to

compute the FWI index. The ISI index indicates the rate of fire spread immediately after

ignition. It combines the FFMC and wind speed to predict the expected rate of fire spread.

Generally, 13 km/h increases in wind speed will double the ISI value. The BUI

index is a weighted combination of the DMC and DC codes, and it indicates the total

amount of fuel available for combustion. The DMC code has the most influence on the BUI

value. For example, a DMC value of zero always results in a BUI value of zero regardless

of what the DC value is. DC has its strongest influence on the BUI at high DMC values,

and the greatest effect that the DC can have is to make the BUI value equal to twice theDMC value. The Fire Weather Index (FWI) is calculated from the ISI and BUI to provide

an estimate of the intensity of a spreading fire. In effect, FWI indicates fire intensity by

combining the rate of fire spread with the amount of fuel being consumed. Fire intensity is

defined as the energy output measured in kilowatts per meter

of flame length at the head of a fire. The head of a fire is the portion of a fire edge showing

the greatest rate of spread and fire intensity.



9/17


The FWI index is useful for determining fire suppression requirements as well as

being used for general public information about fire danger conditions. Although FWI is

not directly calculated from weather data, it depends on those factors through ISI and BUI.

2.3 Interpreting and Using the FWI System

There are two goals of the proposed wireless sensor network for forest fires: (i)

provide early warning of a potential forest fire, and (ii) estimate the scale and intensity of

the fire if it materializes. Both goals are needed to decide on required measures to combat a

forest fire. To achieve these goals, designs the sensor network based on the two main

components of the FWI System: (i) the Fine Fuel Moisture Code (FFMC), and (ii) the Fire

Weather Index (FWI). The FFMC code is used to achieve the first goal and the FWI index

is used to achieve the second. In the following, here it is justified the choice of these two

components by collecting and analyzing data from several forestry research publications.

(a) Probability of ignition as a function (b) Fire intensity as a function of the

FWI index of the FFMC code

Figure 2.3. Using two main components of the Fire Weather Index System in designing a

wireless sensor network to detect and combat forest fires.

The FFMC indicates the relative ease of ignition and flammability of fine fuels due

to exposure to extreme heat. To show this, interpolates data from [de Groot 1998] to plot

the probability of ignition as a function of FFMC. The results are shown in Figure 2.4(a).

The FFMC scale ranges from 0101 and is the only component of the FWI System without



10/17


an open-ended scale. Generally, fires begin to ignite at FFMC values around 70, and the

maximum probable value that will ever be achieved is 96 [de Groot 1998]. Based on data

available from the web site of The Sustainable Resource Development Ministry of the

Province of Alberta, Canada, Table 2 classifies the potential of fire ignition versus the

FFMC ranges. Low values of FFMC are not likely to be fires and can be simply ignored,while larger values indicate more alarming situations.

The FWI index estimates the fire intensity by combining the rate of fire spread

(from the Initial Spread Index, ISI) with the amount of fuel being consumed (from the

Buildup Index, BUI). A high value of the FWI index indicates that in case of fire ignition,

the fire would be difficult to control. This intuition is backed up by several studies. For

example, in 1974, the Alberta Forest Service performed a short term study of experimental

burning in the Jack pine forests in north eastern Alberta. Snapshots of the resulting fires and

the computed FWI indexes are shown in Figure 2.4.

Figure 2.4. Experimental validation of the FWI index.

Another study relates the fire intensity with the FWI index. This relationship is

plotted in Figure 2.3(b) by interpolating data from [de Groot 1998]. In Table.3 provide a

classification of fire danger as a function of the FWI index.

Both the FFMC code and the FWI index are computed from four basic weather

conditions: temperature, relative humidity, precipitation, and wind speed. These weather

conditions can be measured by sensors deployed in the forest. The accuracy and the

distribution of the sensors impact the accuracy of the FFMC code and the FWI index.Therefore, it is needed to quantify the impact of these weather conditions on FFMC and



11/17


FWI. Using this quantification, here designs the wireless sensor network to produce the

desired accuracy in FFMC and FWI.

Table 2. Ignition Potential versus FFMC value.

Ignition Potential FFMC Value Range

Low 0-76Moderate 77-84

High 85-88

Very high 89-91

Extreme high 92+

In addition, this quantification could help other researchers and sensor

manufacturers to customize or develop new products that are more suitable for the forest

fire detection application.

Table 3. Potential Fire Danger versus the FWI value.

FWI ClassValueRange

Type of fire Potential Danger

Low 0-5 Creeping surface fire Fire will be self extinguishing

Moderate 5-10 Low vigor surface fireEasily suppressed with hand

tools

High 10-20Moderately vigorous

surface fire

Power pumps and hoses are

needed

Very high 20-30Very intense surface

fireDifficult to control

Extreme 30+Developing active

fireImmediate and strong action is

critical

Sample of the results are shown in Figure 2.5 and Figure 2.6. The sensitivity of

FFMC to temperature and relative humidity is shown in Figure 2.5 for fixed wind speed at

5km/h and precipitation level of 5mm. Figure 2.6 shows the sensitivity of FWI to

temperature and relative humidity under similar conditions. An interesting observation for

sensor manufacturers is that the accuracy of the sensor readings is critical in high

temperature ranges and when humidity is low, while fine accuracy is not that important

outside these ranges. These figures are used to bound the errors in estimating FFMC and

FWI in the next section. Low values of FFMC are not likely to be fires and may be ignored.

In case of higher FFMC values, where a fire is possible, based on the values of

FWI, some fires might be left to burn, some should be contained and others need to be

extinguished immediately. Here designs the wireless sensor network for forest fire

detection based on the FFMC code and FWI index. This system uses weather data collected

by sensor nodes to calculate these indexes.



12/17


(a) FFMC versus Temperature (b) FFMC versus Humidity

Figure 2.5. Sensitivity of the FFMC code to basic weather conditions.

(a) FWI versus Temperature (b) FWI versus Humidity

Figure 2.6. Sensitivity of the FWI Index to basic weather conditions



13/17


CHAPTER-3

EARLY DETECTION OF FOREST FIRES USING

WIRELESS SENSOR NETWORKS

In this section, this paper presents the design of a wireless sensor network for forest

fire detection. Indeed there are many research problems in such large-scale sensor network.

It focuses on a subset of them, and leverage solutions for other problems in the literature, as

outlined below.

Figure 3.1. Architecture of the proposed forest fire detection system.

The system considered in this paper is depicted in Figure 3.1. A sensor network

deployed in a forest reports its data to a processing center for possible actions, such as

alerting local residents and dispatching fire fighting crews. Sensors are deployed uniformlyat random in the forest by, for example, throwing them from an aircraft. A single forest fire

season is approximately six months (between April and October), and it is desired that the

sensor network lasts for several seasons. Since the lifetime of sensors in active mode is

much shorter than even a fraction of one season, sensor deployment is assumed to be

relatively dense such that each sensor is active only during a short period of time and the

monitoring task is rotated among all sensors to achieve the target network lifetime.

Therefore, during the network operation, a small fraction of the deployed sensors are kept

in active mode, while the rest are put in sleepmode to conserve energy.



14/17


It is important to mention that the forest fire detection application considered in this

report works on a large time scale. Thus, active sensors are not continuously monitoring the

area. Rather, they periodically (e.g., every 30 minutes) perform the sensing task. Therefore,

sensors in the active mode are further divided into active-senseand active-listen

modes. In

the former, all modules (transmission, receiving, and sensing) of the sensor are turned on,

while in the latter only receiving module is on.

Sensors are assumed to self-organize into clusters using a distributed protocol.

After the termination of the clustering protocol, sensors know their cluster heads and the

whole network is connected. Any of the protocols described in the recent survey in

[Youninet al. 2006] can be employed. The proposed system does not restrict them cluster

size, and it allows single- and multi-hop intra-cluster communications. The sensor

clustering and data routing problems are outside the scope of this proposed scheme. This

technology considers four problems. The final problem is achieving unequal fire protection

in different zones in the forest, e.g., forest zones near industrial plants and residential areas,

or forest zones with drier conditions and higher temperatures (denoted by hot spots). This is

illustrated in Figure 3.1 by activating more sensors in the shaded hot spot area.

3.1 Unequal Monitoring of Forest Zones

Unequal monitoring of different forest zones is important in forest fire detection

systems, because some areas may have higher fire potential than others. For example, dry

areas at higher elevations are more susceptible to fires than lower and more humid areas.

Moreover, it is usually important to monitor parts of the forest near residential and

industrial zones with higher reliability and accuracy. To confirm the above intuition, by

collecting real data on the fire danger rating produced by the Protection Program of the

Ministry of Forests and Range, in the Province of British Columbia, Canada. Sample of

sthe data is shown in Figure 3.2 for 23 July 2007. The figure shows several hot spots with

High danger rating within larger areas with Moderate rating. The number, size, and

locations of the hot spots are dynamic, because they depend on weather conditions. Maps

such as the one shown in Figure 3.2 are produced daily.

To support unequal monitoring of forest zones, this technology propose to cover the

forest with different degrees of coverage at different zones. Intuitively, in hot spots, the

FFMC and FWI are expected to be in the high ranges of their scales, and small errors in



15/17


these ranges could lead to misclassifying a fire and/or taking the wrong re-actions. For

example, the Very High range of FFMC in Table II is 8991 (only two units), while the

Low range is 076. Higher accuracy in computing FFMC and FWI require collecting

weather conditions more accurately.

Figure 3.2. Need for coverage with different degrees in forest fires.

Thus it saves the overhead of localization protocols, or the cost of equipping sensors

with GPS, which is a significant saving considering the scale of the forest fire detection

system. However, cluster heads need to determine whether or not they are inside some hot

spots. This can be achieved by associating sensor IDs to their approximate locations during

the deployment process. For example, during deployment, sensors with specific ID ranges

can be thrown by the aircraft in target geographical locations. This mapping is maintained

by the data processing center to dynamically configure the sensor network.



16/17


CHAPTER-4

CONCLUSION

This technology presents the design of a wireless sensor network for early detection

of forest fires. The design is based on the Fire Weather Index (FWI) System, which is

backed by decades of forestry research. The FWI System is comprised of six components:

three fuel codes and three fire indexes. The three fuel codes represent the moisture content

of the organic soil layers of forest floor, whereas the three fire indexes describe the

behavior of fire. By analyzing data collected from forestry research, and showed how the

FWI System can be used to meet the two goals of a wireless sensor network designed for

forest fires:

(i) Provide early warning of a potential forest fire, and

(ii) Estimate the scale and intensity of the fire if it materializes.

To achieve these goals, designed the sensor network based on two main components of the

FWI System: the Fine Fuel Moisture Code (FFMC), and the Fire Weather Index (FWI).

The FFMC code is used to achieve the first goal and the FWI index is used to achieve the

second.



17/17


REFERENCES

[1]. MOHAMED HEFEEDA AND MAJID BAGHERI. Canada Forest Fire Modelling and

Early Detection using Wireless Sensor Networks

[2]. A.PRASHANTH M.Tech, Prof K.Ashok Babu, Sri Indu College of Engg and Tech

Andhra Pradesh, India Forest Fire Prevention using Wireless Sensor Networks.

[3]. KRESIMIR PRIPUZIC, HRVOJE BELANI, MARIN VUKOVIC. Croatia. Early

Forest Fire Detection with Sensor Networks: Sliding Window Skylines Approach.

[4]. LIYANG YU, NENG WANG, East China Normal University,China. Real-time Forest

Fire Detection with Wireless Sensor Networks.

Dept of ECE CIT Gubbi 17

forest fire prevetion

Documents